Manga colorization and anime generation
This project aims to adapt state-of-the-art diffusion models for text-assisted, consistent manga colorization. It involves developing both a novel approach for evaluating colorization consistency and an interactive tool that enables user-guided colorization workflows.
Tasks:
Data annotation for image captioning
Data annotation for character reidentification
Colorization tool development
Task Descriptions
Data annotation for image captioning - Polish English captions for manga panels generated by a VLM
Skills:
Computer proficiency
English B2
Data annotation for character reidentification - Polish character annotation (bbox/segm + id) for manga pages generated by a model
Skills:
Computer proficiency
English/Japanese/Chinese proficiency
Colorization tool development - Develop web tool for interactive manga colorization
Skills:
Computer proficiency
English/Japanese/Chinese proficiency
Participants
Ilya Makarov
Team lead
Maxim Golyadkin
Project lead
Related Publications
Closing the Domain Gap in Manga Colorization via Aligned Paired Dataset
IEEE/CVF Winter Conference on Applications of Computer Vision (WACV) (2025)
Maxim Golyadkin, Ilya Makarov